List your company

Tool Use on RobustBench-TC Perturbed 1.0 (test)

52.9Accuracy (Perturbed)Qwen3-14B
25.65232.72639.846.874May 12, 2026

Evaluation results

MethodLinks
Qwen3-14BModel Category=Base / instruct counterparts, Model Scale=14B2026.0552.90.0390.1430.4640.25
MUA-RL-32BModel Category=Open-source RL tool models, Model Scale=32B2026.0552.70.028-0.010.3970.218
Qwen3-32BModel Category=Base / instruct counterparts, Model Scale=32B2026.0552.40.0240.1040.3660.377
LoopTool-32BModel Category=Open-source RL tool models, Model Scale=32B2026.0550.80.0340.120.5090.396
o4-mini (OpenAI)Model Category=Closed-source frontier2026.0550.20.0490.1570.3260.276
MUA-RL-8BModel Category=Open-source RL tool models, Model Scale=8B2026.0547.90.0230.1560.4010.229
LoopTool-8BModel Category=Open-source RL tool models, Model Scale=8B2026.0547.60.0490.1730.4290.296
MUA-RL-14BModel Category=Open-source RL tool models, Model Scale=14B2026.05470.0110.090.3340.202
ToolRL-DR-Full (ours)Model Category=Our 3B trained checkpoints, Model Scale=3B2026.0546.30.0090.1470.3310.231
ToolRL-DR-Mixed (ours)Model Category=Our 3B trained checkpoints, Model Scale=3B2026.0546.10.0490.1520.3640.232
Qwen3-8BModel Category=Base / instruct counterparts, Model Scale=8B2026.0544.80.0190.1710.4090.293
DeepSeek-R1-Distill-14BModel Category=Frontier, Model Scale=14B2026.0544.20.0130.160.3180.213
Qwen3.5-9BModel Category=Frontier, Model Scale=9B2026.0543.70.040.1680.4110.417
ToolRL-Qwen2.5-1.5BModel Category=Open-source RL tool models, Model Scale=1.5B2026.0543.20.0330.1740.4810.379
TL-CodeLLaMA-2Model Category=Open-source RL tool models, Model Scale=7B2026.0540.50.0240.1660.4560.333
ToolRL-Qwen2.5-3BModel Category=Open-source RL tool models, Model Scale=3B2026.0540.40.020.1430.450.315
ToolRL-Llama3.2-3BModel Category=Open-source RL tool models, Model Scale=3B2026.05390.0250.1710.4630.296
Qwen2.5-3B-InstructModel Category=Base / instruct counterparts, Model Scale=3B2026.05390.0140.1950.4460.237
Qwen2.5-1.5B-InstructModel Category=Base / instruct counterparts, Model Scale=1.5B2026.0536.70.040.20.4050.205
Llama-3.2-3B-InstructModel Category=Base / instruct counterparts, Model Scale=3B2026.0534.10.0090.1770.3720.174
SFT-Clean-4k-3BModel Category=ToolRL SFT version, Model Scale=3B2026.0526.70.0190.1840.4430.577